~~~
Due to Segwit the blocksize isn't limited to 1MB only. The current average blocksize is around 1.7MB, see e.g.
https://ycharts.com/indicators/bitcoin_average_block_size. Therefore yearly growth is noticeably larger than your calculated ~51.36GB. With a "Segwit factor" of around 1.7 we'd land at about ~87.3GB which is pretty close to what I got from a chart.
My earlier written value of around 91GB (not exactly sure if it's actually 91GiB) is based on values from a graph which displayed total blockchain size in GB over the last two years. The chart had a slight rise in the growth slope around beginning of 2023, so I didn't use a larger observance window but rather the slightly worse growth rate of last two years (reason for worse growth rate is debatable).
I then simply calculated 200y times 91GB/y neglecting possible growth of transaction indices and UTXO set (chainstate). People send daily numerous coin pieces mostly to the Genesis block which bloats the UTXO set as I don't expect Satoshi to consolidate UTXOs "donated" to the Genesis block or block #9.
A larger UTXO set puts a burden on RAM need for a speedy IBD as it's beneficial to keep the UTXO set in RAM during IBD otherwise speed suffers from heavy I/O to storage media (will be painfully slow with low random IOps devices like mechanical harddrives).
I'm not too worried about the size and growth of the UTXO set over time.
Do you think the size of the blockchain will bloat more than my calculation suggests just 20 years from now?
You proposed a number after 200 years where you didn't explain how it was calculated and which I highly doubt to be any close. Why 200 years anyway? It's already hard to make predictions of technological advances for the next few decades.
As we don't know how Bitcoin will evolve, I find it already a bit difficult to make more or less accurate predictions for the next few decades.
I started by asking whether the continuously growing blockchain would become a burden on nodes in the future.
So far, I think it's managable and as computers evolve, too, I don't see yet a real problem.
My last experiment with a low power device, a Raspi 4B with 8GiB RAM, in June 2023 finished a full IBD from scratch up-to block 796033 in about 95h with a SATA 1TB SSD connected via an USB3-SATA adapter, network connection was Tor-only with a stable 100MBit internet connection. I expected it would take longer and was positively surprised.